Architectural Styles and Quality Attributes in AI-Based Video Software: A Systematic Literature Review

AI-based video software systems are increasingly prevalent in surveillance, content generation, and multimedia analytics domains. However, we know little about these systems’ architectures or how they address software-level quality attributes. This gap in documentation limits the reproducibility of such systems. This paper addresses this issue and presents a Systematic Literature Review (SLR) of 139 peer-reviewed studies published between 2020 and 2025. It investigates which software architecture styles they report, the reasons behind their selection, and the quality attributes evaluated in developing AI-based video systems. The results show that only eight studies (≈6%) mention a software architecture style—primarily client-server, component-based, cloud-based, and serverless—and only 10 (≈7%) explicitly discuss software quality attributes, such as efficiency, complexity, robustness, performance, and scalability. The findings highlight a significant literature gap in architectural and quality-related reporting. This review contributes new insights into underexplored aspects of AI-based video systems by elucidating the interplay between architecture decisions and system-level quality outcomes. It emphasizes the need for more systematic documentation of software engineering decisions to support reliable, maintainable, and scalable AI applications.